Sei sulla pagina 1di 9

Int J Adv Manuf Technol (2009) 45:481489

DOI 10.1007/s00170-009-1984-0

ORIGINAL ARTICLE

Three-dimensional modeling of roughness effects


on microthickness filling in injection mold cavity
Nan S. Ong & Hong L. Zhang & Yee C. Lam

Received: 8 October 2008 / Accepted: 16 February 2009 / Published online: 7 March 2009
# Springer-Verlag London Limited 2009

Abstract Effects of cavity surface roughness on filling


polymer flow in macrocavities are generally not significant.
However, in a microcavity, surface roughness may become
very important and ignoring it could lead to inaccurate or
even misleading results. As an extension to our previous
work, a three-dimensional roughness model for filling
polymer flow in microinjection molding is proposed that
takes into consideration the cavity surface roughness
effects. A numerical procedure, which implements the
finite volume and level set methods and integrates the
proposed roughness model, is presented. The numerical
model was validated by comparing its predicted results with
experimental and/or analytical results. The numerical
results obtained using the three-dimensional roughness
model were in good agreement with experimental observations and were more accurate than those obtained using the
two-dimensional roughness model.
Keywords Mold roughness . Microcavity filling .
Numerical model . Level set

1 Introduction
The effects of cavity surface roughness on filling polymer
flow are generally not significant for a macrocavity (e.g.,
the thickness of the cavity is greater than 1 mm). However,

N. S. Ong (*) : H. L. Zhang : Y. C. Lam


School of Mechanical and Aerospace Engineering,
Nanyang Technological University,
Nanyang Avenue,
Singapore 639798, Republic of Singapore
e-mail: mnsong@ntu.edu.sg

when the thickness of a cavity is decreased until the


roughness is comparable to the dimension of the cavity, the
effects of surface roughness become increasingly important.
It may significantly change the physical volume of the
cavity and enhance heat transfer between the polymer flow
and the cavity [1]. Therefore, ignoring surface roughness in
the analysis of filling polymer flow in a microcavity may
lead to significantly inaccurate results.
Among the existing investigations of roughness effects
on fluid flow in microchannels, a few researchers [2, 3]
considered the roughness zone as a porous medium layer
(PML) and the analysis was performed based on the
available PML theory. Others [4, 5] took into account the
roughness by using roughnessviscosity models. These
models, however, are not suitable for analyzing roughness
effects on filling polymer flow. For example, in the PML
approach, the volume porosity and permeability of the
roughness layer are difficult to determine in practice.
Moreover, these models were mainly used for Newtonian
and isothermal fluid flow, without considering the interaction between the flow field and the thermal field, which,
however, is predominant in microinjection molding.
Recently, we [6] proposed a two-dimensional roughness
model, using two-dimensional homogenous roughness
profile, such as homogenous rectangular or triangular
roughness elements, to represent the physical roughness
texture and thermal properties of the roughness zone. The
thermal conductivity and heat capacity are modeled based
on a simple rule of mixture. The numerical results obtained
using the two-dimensional roughness model are in acceptable agreement with the experimental data. However,
roughness is inherently three-dimensional. Thus, when the
two-dimensional roughness model is used, the full threedimensional roughness texture, such as hills that repeat in
both directions, is not taken into consideration. This may

482

lead to inaccurate prediction of the roughness effects on


polymer flow in the cavity.
Thus, we propose here to adopt a three-dimensional
homogenous pyramid roughness profile to represent the
physical roughness texture. A numerical procedure, which
implements the finite volume and level set methods that
integrates the proposed roughness model, is presented. The
numerical model was validated by comparing its predicted
results with experimental and/or analytical results. The
numerical results obtained using the three-dimensional
model were compared with the experimental data and with
those obtained using the two-dimensional roughness model.

2 Experimental design
In our previous work [7], injection molding of a thin
circular disk with polyoxymethylene (POM) was performed
on an injection molding machine. As shown in Fig. 1, the
wall of the cavity insert has a smooth surface while the wall
of the core insert has different roughness on its two
semicircular halves. Correspondingly, the top of the plastic
part has a relatively smooth surface, whereas the bottom
has different roughness on its two semicircular halves.
In the experiments, the separating line between the
smoother and the rougher halves of the core wall was
adjusted vertically in the mold such that gravity has the
same effect on filling the two semicircular cavity halves.
Therefore, the two semicircular halves of the cavity can be
assumed to be filled under the same process conditions (i.e.,
the same mold and melt temperatures, filling pressure, and
injection rate) and the difference in the flow areas between
the two halves of the molded part was predominantly
caused by the difference in surface roughness between the
two semicircular halves of the core wall.

Fig. 1 Schematic diagram of cavity filling encountered in the


experiments

Int J Adv Manuf Technol (2009) 45:481489

3 Three-dimensional representation of surface


roughness
A three-dimensional homogeneous pyramid roughness
profile over a surface as shown in Fig. 2 is proposed to
represent the physical roughness texture. The threedimensional average roughness, Sa, of the pyramid roughness profile is defined as:
Sa

1
3

AB d 13 AC d
; AB AC
AB AC

where, AB and AC, which are assumed to be the same, are


areas at the base of the pyramid roughness above and below
the mean plane, respectively. is the effective roughness
height (i.e., the distance from the roughness peaks to the
roughness mean plane) and =3Sa. The average roughness
Sa of a physical surface can be easily obtained through
experimental measurement.For the pyramid roughness
profile where AB =AC, in any imaginary cross-sectional
plane of y, the area occupied by the mold material (AD) is:


dy 2
AD
AB
2
d
Thus, the volume fraction () occupied by polymer flow on
the plane of y can be obtained as:
(
 2
1  12 dy
0yd
q
:
3
  d
1 dy 2
d  y < 0
d
2
Similar to the two-dimensional roughness model as
proposed previously [6], the thermal conductivity within
the roughness layer (see Fig. 3) is modeled as:
k y q yk1 1  q yk2

where, k1 and k2 are the thermal conductivities of the


polymer melt and the mold material, respectively, and is
the volume fraction occupied by the polymer melt at any

Fig. 2 Three-dimensional homogeneous pyramid roughness profile

Int J Adv Manuf Technol (2009) 45:481489

483

Fig. 3 Schematic diagram for


determination of thermal properties of roughness layer

cross-sectional plane, which is parallel to the roughness


mean plane.
Similarly, the heat capacity (in terms of density and the
specific heat) of the roughness layer is modeled as:
rCP y q yr1 CP1 1  q yr2 CP2

where, 1CP1 and 2CP2 are the heat capacities of the melt
and the mold, respectively.
By definition, the cross-sectional area/volume subtended
by the roughness profile above and below the mean plane
has equal value. Thus, the effective wall (see Fig. 3), which
is defined to be the same as the mean plane, should be used
in the analysis. This is to ensure that the volume enclosed
between the two effective walls (mean planes) of a cavity
corresponds to the volume of the physical cavity.

@
~
v  r 0 ; jrj 1
@t

4 Numerical model
To simulate the filling polymer flow as encountered in the
experiments, a numerical model was developed. During the
filling of polymer melt into the cavity, gravity can be
neglected as the gravitational force is negligible as compared
to the viscous force of the melt [8]. Surface tension can also
be neglected as the dimensions of the molded parts are
much larger than 1m [9]. As a result, the governing
equations for the flow in the cavity may be written as:
@r
r  r~
v 0
@t


g r~
v r~
v



 
@r~
v
rr~
v~
v rp r  hg ;
@t

@rCP T
~
v  rrCP T
@t
r
1  

g:g
g
2

where, , , Cp, and are the density, viscosity, specific


heat, and conductivity of the polymer melt, respectively.
However, for energy Eq. (8), in the roughness layer
occupied by the polymer melt, and CP are to be
modified and determined using the proposed roughness
model. In the momentum and energy equations (i.e., Eqs.
(7) and (8)), the cross-WLF model [10] is employed to
describe the polymer viscosity, . Moreover, the variation
in polymer density with temperature is taken into account
using the modified two-domain Tait PVT model [11].
The flow front of a polymer melt is captured using the
level set method [12]. A scalar variable is used to identify
the interface between polymer melt and air, i.e., the zerolevel set of indicates the polymerair interface, <0
indicates the polymer occupied region and <0 indicates
the air occupied region. The transport equation of the level
set function, , is written as

r  krT hg 2 ;
8

To maintain as a distance function at all time steps, a


re-distancing (or re-initializing) step is adopted (see Eq.
(10)). Moreover, a mass correction procedure developed by
Yap et al. [13] is used to ensure global mass conservation.
@x
sign1  jrxj; x~
x; 0 ~
x; t Dt
@t 0

10

The high ratio of the properties between polymer melt


and air, such as density, viscosity, thermal conductivity,
etc., tends to cause instability and divergence of the
numerical solution. To overcome this difficulty, and with
negligible sacrifice of accuracy, the fictitious fluid approach
[8, 14] was employed. In this approach, a fictitious fluid is
introduced to represent the air downstream of the polymer
flow front. The main consideration of this fictitious fluid is
that its viscosity is at least 103 times smaller than that of
the polymer melt, exceeding the real value of air viscosity
by several orders of magnitude. The other properties of the
fictitious fluid are the same as those of air. Moreover, a
Heaviside function is adopted to define a smeared interface

484

Int J Adv Manuf Technol (2009) 45:481489

Table 1 Analytical and predicted results of the Poiseuille number (fRe)


0

486.33
381.63
302.87

17,795.55
15,895.24
15,103.00

0.41
0.42
0.42

Analytical Re (Va)

Predicted fRe (Vb)




a
Simulation error VbVV
  100%
a

44.9
46.5
47.8

45.5
46.9
48.2

1.3%
0.9%
0.8%

between polymer melt and the fictitious fluid [15], i.e., the
property near the interface is calculated using
a 1  ha1 ha2

11

where, 1 and 2 stand for the properties of polymer melt


and the fictitious fluid, respectively, such as density,
viscosity, specific heat, and thermal conductivity. The
Heaviside function h is written as:
8
<0
 p  < "
1
h "
12

sin
jj  " :
2p
"
: 2"
1
>"
where, is related to the grid size and is usually taken as a
factor of the grid spacing. A thickness of 2 is recommended.
According to the proposed roughness model, the
physical roughness profile is replaced by a homogenous,
uniform pyramid roughness profile. Noting that the x and z
dimensions of the surface roughness is many orders smaller
than the cavity, the detail local xz variation of the profile
could be ignored. Instead, the average of its effects on heat
capacity and conductivity could be calculated. With this
approximation, the thermal conductivity and heat capacity
of the roughness layer are isotropic in the radical and

Fig. 4 Experimental, analytical, and simulation results of average


flow front position versus filling time

circumferential directions and vary only in the thickness


direction. With this simplification, the physically threedimensional roughness layer could be considered as twodimension in the simulation.
However, the filling flow is not strictly axisymmetric
due to the difference in roughness between the two
semicircular halves of the core insert. To simulate such a
non-axisymmetric (or pseudo-axisymmetric) filling flow,
the inletvelocity iteration approach developed was
employed here [16]. In this approach, the flow on either
half is assumed to be axisymmetric. It is noted that for the
two halves, (a) the applied pressure has to be the same, and
(b) the volume flow rates are different, but the sum of the
volume flow rates must equal to the total volume flow rate.
Thus, during each calculation step, the inlet velocities on
the two semicircular halves change simultaneously and in
opposite directions to maintain the required total volume
flow rate through iterations. Iteration is terminated when
the pressures at the center of the core insert for the two
halves can be considered the same within the pre-defined
tolerance range.
At the inlet, the velocity and temperature of polymer
melt are kept constant. At the outlet, outflow boundary
condition is imposed. The temperature of the melt/mold
interface is assumed to be equal to the wall temperature,
namely, the heat transfer coefficient between the mold and
the melt is assumed to be very high. Although wall slip
may occur for some fluids in certain conditions and may be
important for gas-assisted injection molding [1719], wall
slip along a rough wall for a polymer flow is not
convincingly observed through experiments [20, 21]. There
is evidence to suggest that a rough wall could prevent wall
slippage [21]. As we did not observe any evidence of wall
slippage in our molded samples, we therefore imposed in
our simulation non-slip boundary condition on the meltfilled domain. However, to allow for the occurrence of the
fountain effect and advancement of the melt flow front, the
free-slip boundary condition is imposed on the unfilled
domain [8, 14].
The partial differential equations as described above are
solved using the finite volume method of Partankar [22]. A
fixed and staggered grid was used, with the scalar variables
stored in the centers of the control volumes, while the
velocities were located at the control volume faces. The

Int J Adv Manuf Technol (2009) 45:481489

485

Fig. 5 Pressure sensor and core


insert assembly

coupling between velocity and pressure was handled by the


SIMPLER algorithm, and the diffusionconvection effect in
the momentum equations was modeled by the power law
scheme. The fully implicit scheme was used to discretize
the transient term. The upwind scheme was used to model
the convection of the level set equations. The resulting
algebraic equations were solved using the TriDiagonal
Matrix Algorithm.

5 Validation of numerical model


In the simulation, the material data of POM (e.g., density,
viscosity, etc.), the coefficients in the modified two-domain

Fig. 6 Schematic connection diagram of the cavity pressure acquisition system

Tait PVT model and in the WLF viscosity model were


obtained from Moldflow MPI 5.0 material database. The
fictitious fluid's density, thermal conductivity, and specific
heat are 1.0 kg/m3, 0.037 W/m K and 1.0 J/kg K,
respectively. The mold material's density, thermal conductivity, and specific heat are 7,800 kg/m3, 29 W/m K, and
460 J/kg K, respectively.
For a preliminary validation, simulations of fully
developed flows of Newtonian and non-Newtonian modified power law (MPL) fluids in an axisymmetric geometry
were performed. For a fully developed laminar flow of
Newtonian fluid in an axisymmetric geometry, the numerically predicted Poiseuille number (i.e., the product of the
friction factor (f) and the Reynolds number (Re)) is 63.96,
which is close to its theoretical value of 64. For fully
developed non-Newtonian MPL flows in an axisymmetric
geometry, as shown in Table 1, the numerical results of the
Poiseuille number also agree well with the analytical ones
[4], with a maximum simulation error of 1.3%.
Numerical results of filling flow length were also
compared with the experimental and analytical results. To
achieve different flow front positions during cavity filling,
short shot experiments were performed through varying
injection pressures and dosages while keeping the injection
rate constant at 3.0 cm3/s. The average position of the melt
front in the radial direction at time t can be obtained
analytically based on the mass conservation principle,
which is given as:
s
2 H
2
dini
eff d1 Uin t
r t
13
4Heff
where, dini, Heff, d1, Uin, and t are, respectively, the initial
location of polymer melt flow in the cavity, effective cavity

486
Table 2 Part list of the cavity
pressure acquisition system

Int J Adv Manuf Technol (2009) 45:481489


No.

Description

Model

Quantity

1
2
3
4
5
6

NI-A/D card
Connecting block
Junction box
Miniature charge amplifier
Quartz miniature force sensor
Cavity pressure sensor

National Instruments PCI-6023E


National Instruments CB-68LP

1
1
1
1
1
1

Kistler 5039A232
Kistler 9211
Kistler 6183AE

Fig. 9 Numerically predicted and experimentally measured cavity


pressure curves for Tmelt =453 K and Tmold =383 K
Fig. 7 Pressure distribution over the cavity pressure sensor

Fig. 8 Numerically predicted and experimentally measured cavity


pressure curves for Tmelt =453 K and Tmold =323 K

Fig. 10 Numerically predicted and experimentally measured cavity


pressure curves for Tmelt =473 K and Tmold =323 K

Int J Adv Manuf Technol (2009) 45:481489

Fig. 11 Numerically predicted and experimentally measured cavity


pressure curves for Tmelt =473 K and Tmold =383 K

thickness, diameter of inlet, injection velocity, and filling


time.
Figure 4 shows the experimental, analytical, and simulation results of the average flow front position versus
filling time. It shows that the simulation result is in good
agreement with both the experimental and the analytical
results.
Subsequently, the numerical procedure was further
validated by comparing its numerically predicted results
of cavity pressure versus time with the experimentally
measured data. To measure cavity pressure, one piezoelectric sensor of 1 mm diameter is mounted inside the core
insert as shown in Fig. 5. The head of the sensor is located
in the center of the core insert. A data acquisition system
was designed to acquire the measured pressure data.
Figure 6 shows the schematic connection diagram of the
pressure acquisition system and Table 2 lists the parts used
in the data acquisition system. The measured pressure is
converted into electric charge, then amplified and converted
into a voltage signal. This signal is finally registered by the
computer through the NI-A/D card and visualized by
DaisyLab software.
As the cavity pressure sensor measures the average
pressure over an area of 1 mm diameter as illustrated in
Fig. 7, in the simulation, the corresponding average

Table 3 Values of Sa/Ra, ,


and for two-dimensional and
three-dimensional roughness
models

Roughness model
Three-dimensional pyramid
Two-dimensional triangle

487

Fig. 12 Experimental and numerical results of flow area of the


rougher half versus the total flow area of molded parts at melt
temperature of 453 K and mold temperature of 323 K

pressure (Pave) is calculated according to the following


expression:
RA

pdA

Pave R A
0

dA

R 0:5

p2prdr

R 0:5
0

2prdr

R 0:5

p2prdr

14

pr2 0:5
0

Figures 811 show the numerically predicted and


experimentally measured cavity pressures versus filling
time for different melt and mold temperatures. The cavity
thickness was 0.36 mm and the injection rate was 2.8 cm3/s.
For each combination of the melt and mold temperatures,
three cavity pressure curves were experimentally obtained.
Reasonable agreement was observed, with the numerically
predicted pressure curves lying within or close to the band
formed by the experimental curves.

6 Validation of the three-dimensional roughness model


To validate the proposed roughness model, numerical
simulation of the pseudo-axisymmetric filling flow encountered in the experiments for a cavity with a rough half were
performed. The numerically predicted flow area development on the rougher half versus the increasing total flow
area was compared with the experimentally measured data.

Sa/Ra (m)

(m)

16.7
10.0

50
20

 2
q 1  12 dy
d
0.5y/+0.5

488

Int J Adv Manuf Technol (2009) 45:481489

Fig. 13 Experimental and numerical results of flow area of the


rougher half versus the total flow area of molded parts at melt
temperature of 453 K and mold temperature of 383 K

Fig. 15 Experimental and numerical results of flow area of the


rougher half versus the total flow area of molded parts at melt
temperature of 473 K and mold temperature of 383 K

In the simulation, the three-dimensional homogeneous


pyramid roughness and the two-dimensional homogenous
triangle roughness were used to represent the physical
roughness of the core inserts. Considering that the
roughness of the smoother half of the core insert is much
smaller than 1 m, its roughness is ignored in the
simulation, and only the roughness on the rougher half is
considered. For the three-dimensional pyramid roughness
model, =3Sa. For the two-dimensional roughness model,
as analyzed in our previous work [6], =2Ra. Table 3
summarizes the values of Sa/Ra, and for the twodimensional and three-dimensional roughness models for
the same physical surface of the rougher half of the cavity.

The diameters of the inlet and the cavity and the melt and
mold temperatures employed in the simulations were the
same as those used in the experiments. The injection rate
was 2.5 cm3/s, which is similar to that used in the
experiments.
Figures 1215 show the experimentally measured and
numerically predicted flow area of the rougher half versus
the increasing total flow area for different melt and mold
temperatures. It can be observed that the numerical results
using the three-dimensional model are in good agreement
with the experimental data. Furthermore, using the threedimensional model provides better agreement than using
the two-dimensional roughness model.

7 Conclusion

Fig. 14 Experimental and numerical results of flow area of the


rougher half versus the total flow area of molded parts at melt
temperature of 473 K and mold temperature of 323 K

In this investigation, a three-dimensional roughness model


was proposed that takes into consideration the roughness
effects on the filling polymer flow in microinjection
molding. The physical roughness elements were represented by homogenous pyramid roughness and the thermal
properties of the roughness region were modeled based on a
simple rule of mixture. The effective height of the
roughness region can be obtained indirectly through
measurement of the commonly used roughness parameter
Sa, which means that the model is convenient for use in
practical applications. The numerical results obtained using
the three-dimensional roughness model were in good
agreement with the experimental data. Furthermore, the
three-dimensional model developed gave a better agreement with the experimental data than the previously
proposed two-dimensional roughness model.

Int J Adv Manuf Technol (2009) 45:481489


Acknowledgements The authors gratefully acknowledge the funding of this research provided by Nanyang Technological University
and the Moldflow Pty. Ltd.

References
1. Theilade UR, Kjaer EM, Hansen HN (2003) The effect of mold
surface topography on plastic part in-process shrinkage in
injection molding. ANTEC 1:463467
2. Koo J, Kleinstreuer C (2003) Liquid flow in microchannels:
experimental observations and computational analysis of microfluidics effects. J Micromechanics Microengineering 13:568579
3. Kleinstreuer C, Koo J (2004) Computational analysis of wall roughness effects for liquid flow in micro-conduits. J Fluid Eng 126:19
4. Mala G, Li D (1999) Flow characteristics of water in microtubes.
Int J Heat Fluid Flow 20(2):142148
5. Qu WL, Mala GH, Li D (2000) Heat transfer for water flow in trapezoidal silicon microchannels. Int J Heat Mass Transfer 43:39253936
6. Zhang HL, Ong NS, Lam YC (2007) Effects of mold surface roughness on micro injection molding. Polym Eng Sci 47(12):20122019
7. Zhang HL, Ong NS, Lam YC (2008) Experimental investigation
of key parameters on the effects of cavity surface roughness in
micro injection molding. Polym Eng Sci 48(3):490495
8. Haagh G, Vosse F (1998) Simulation of three-dimensional
polymer mould filling processes using a pseudo-concentration
method. Int J Numer Methods Fluids 28:13551369
9. Yao DG, Kim B (2002) Simulation of the filling process in micro
channels for polymeric materials. J Micromechanics Microengineering 12:604610
10. Koh YH, Ong NS, Chen X, Lam YC, Chai JC (2004) Effect of
temperature and inlet velocity on the flow of a nonNewtonian
fluid. Int Commun Heat Mass Transf 31(7):10051013

489
11. Chen X, Bhagavatula N, Castro JM (2004) Predicting fill patterns
for the in-mould coating process for thermoplastic parts. Model
Simul Mater Sci Eng 12:267287
12. Osher S, Sethian A (1988) Fronts propagating with curvaturedependent speed: algorithms based on HamiltonJacobi formulations. J Comput Phys 79:249
13. Yap YF, Chai JC, Wong TN, Toh KC, Zheng HY (2006) A global
mass correction scheme for the level-set method. Numer Heat
Transf 50(5):455472
14. Hetu JF, Gao DM, Rejon AG, Salloum G (1998) 3D finite element
method for the simulation of the filling stage in injection
moulding. Polym Eng Sci 38(2):223236
15. Chang YC, Hou TY, Merriman B, Osher S (1996) A level set
formulation of Eulerian interface capturing methods for incompressible fluid flows. J Comput Phys 124:449464
16. Zhang, H.L. (2008) Numerical and Experimental Investigation on
Cavity Roughness Effects in Micro Injection Molding. PhD
Thesis, School of Mechanical and Aerospace Engineering, NTU,
Singapore
17. Miksis MJ, Davis SH (1994) Slip over rough and coated surfaces.
J Fluid Mech 273:125139
18. Jansons KM (1985) Moving contact lines on a two-dimensional
rough surface. J Fluid Mech 154:128
19. Dimakopoulos Y, Tsamopoulos J (2006) Gas-assisted injection
molding with fluids partially occupying straight or complex tubes.
Polym Eng Sci 46(1):4768
20. Griffiths CA, Dimov SS, Brousseau EB, Hoyle RT (2007) The
effects of tool surface quality in micro-injection moulding. J Mater
Process Technol 189:418427
21. Smialek CD, Simpson CL (1998) Flow instabilities in thin-wall
injection molding of thermoplastic polyurethane. ANTEC
3:33733377
22. Patankar SV (1980) Numerical heat transfer and fluid flow.
Hemisphere, Washington DC

Potrebbero piacerti anche